RemoteCare: AI-Driven Multimodal Predictive Framework With Blockchain for Personalized Remote Patient Monitoring in IoMT

Abstract

The Internet of Medical Things (IoMT) enables continuous health monitoring but still faces challenges in achieving personalized predictions and ensuring secure, tamper-proof data integrity. We present RemoteCare, an AI-driven multimodal framework that fuses synchronized physiological and network data for dual-task learning, simultaneously performing personalized health state classification (normal, warning, and critical) and cyberattack detection in IoMT traffic. Unlike conventional population-based thresholds, RemoteCare dynamically adapts alerts to each patient’s baseline, thereby minimizing false alarms and enhancing clinical reliability. A hybrid convolutional neural network (CNN)–gated recurrent unit (GRU)–long short-term memory (LSTM) architecture jointly captures spatial and temporal dependencies across heterogeneous signals, while Shapley additive explanations (SHAPs)-based explainability provides transparent, patient-specific insights into the features influencing each prediction. To guarantee auditability, all predictions are immutably recorded on the PureChain blockchain integrated with interplanetary file system (IPFS), ensuring decentralized and tamper-proof storage. Evaluated on the WUSTL-EHMS-2020 dataset (enhanced healthcare monitoring system), RemoteCare achieved 99.7% accuracy for health classification and 96.0% for intrusion detection, with negligible false alarms and efficient inference suitable for real-time deployment. By unifying multimodal prediction, personalization, interpretability, and secure logging, RemoteCare establishes a trustworthy framework for early intervention, patient-specific risk assessment, and clinician-oriented decision support in remote healthcare.